Disease-Specific Differences in Pharmacokinetics of Paromomycin and Miltefosine Between Post-Kala-Azar Dermal Leishmaniasis and Visceral Leishmaniasis Patients in Eastern Africa

Abstract

Treatment regimens for post-kala-azar dermal leishmaniasis (PKDL) are usually extrapolated from those for visceral leishmaniasis (VL), but drug pharmacokinetics (PK) can differ due to disease-specific variations in absorption, distribution, and elimination. This study characterized PK differences in paromomycin and miltefosine between 109 PKDL and 264 VL patients from Eastern Africa. VL patients showed 0.55-fold (95% confidence interval [CI], .41–.74) lower capacity for paromomycin saturable reabsorption in renal tubules, and required a 1.44-fold (95% CI, 1.23–1.71) adjustment when relating renal clearance to creatinine-based estimated glomerular filtration rate. Miltefosine bioavailability in VL patients was lowered by 69% (95% CI, 62%–76%) at treatment start. Comparing PKDL to VL patients on the same regimen, paromomycin plasma exposures were 0.74- to 0.87-fold, while miltefosine exposure until the end of treatment day was 1.4-fold. These pronounced PK differences between PKDL and VL patients in Eastern Africa highlight the challenges of directly extrapolating dosing regimens from one leishmaniasis presentation to another.

Significant pharmacokinetic differences hinder direct extrapolation of dose regimens from visceral leishmaniasis (VL) to post-kala-azar dermal leishmaniasis (PKDL) patients. PKDL patients on a similar regimen as VL patients showed reduced paromomycin exposure and increased miltefosine exposure.

Visceral leishmaniasis (VL), also known as kala-azar, is the most severe form of the neglected tropical parasitic disease leishmaniasis and is fatal if left untreated [1]. Post-kala-azar dermal leishmaniasis (PKDL) is a complication of VL that appears as skin lesions months or even years after successful treatment of a primary VL infection [2]. The infected skin of PKDL patients can serve as a reservoir for parasites [3, 4]. As a result, PKDL is believed to play a pivotal role in the transmission of VL, making the treatment of PKDL patients a crucial element in efforts to eliminate VL.

In Eastern Africa, where VL is caused by Leishmania donovani, the current standard of care is a 17-day combination therapy of sodium stibogluconate and paromomycin (SSG/PM) [5, 6]. A recent phase 3 study demonstrated that a 14-day combination treatment of paromomycin and miltefosine shows similar efficacy as SSG/PM, suggesting its potential to replace the toxic SSG-based regimen [7]. Notably, within 6 months of successful SSG/PM treatment, 21% of VL cases develop PKDL in Sudan and Ethiopia [7]. There is no consensus on the optimal treatment for PKDL. In Sudan, the current practice is treatment with SSG for >60 days, which is associated with life-threatening toxicities and requires long hospitalization [8]. In a recent clinical study on PKDL in Sudan, a short course of liposomal amphotericin B or paromomycin was combined with prolonged oral miltefosine, requiring shorter hospitalization while maintaining high efficacy [9].

The dosing regimen and rationale for treatment of PKDL is typically extrapolated from those used for VL, with an extended treatment period due to the slow recovery of skin lesions [2]. However, VL and PKDL patients exhibit marked differences in disease presentation and physiological status. In VL, parasites manifest mainly in macrophages in spleen, liver, and bone marrow, while in PKDL, the parasites reside locally in the skin lesions. As a result, VL patients are systemically severely ill and usually suffer from fever, weight loss, and hematological depletions, whereas PKDL patients generally maintain better overall health except for their skin condition. These distinctions could potentially result in variations in the pharmacokinetics (PK) of antileishmanial drugs and the relationships between drug exposure and response.

The PK of paromomycin and miltefosine have been studied in VL patients, yet no PK data were previously available for PKDL patients. Paromomycin, a highly hydrophilic aminoglycoside, is rapidly absorbed following intramuscular injection and primarily excreted unchanged via glomerular filtration [10, 11]. Renal function, expected to be better in PKDL compared to VL patients [12], may influence the clearance of paromomycin. Previous population PK studies in VL patients demonstrated changes in miltefosine bioavailability over a 4-week treatment period, with a decrease of 69%–74% in the first week, probably due to malabsorption related to the systemic infection and/or malnourishment [13, 14]. Whether this effect is specific to VL patients or also applies to PKDL patients is unknown. Consequently, directly applying dosing regimens from VL to PKDL patients may lead to bias, with potential dose adjustments required for PKDL patients.

To optimize dosing in PKDL patients, understanding the PK differences between PKDL and VL patients is crucial. This study aimed to characterize the systemic PK of paromomycin and miltefosine in PKDL and VL patients from eastern Africa and to identify disease-specific factors affecting PK.

METHODS

Clinical Trials and Patients

Pharmacokinetic data were collected from a PKDL (NCT03399955) [9] and a VL clinical study (NCT03129646) [7] in Eastern Africa. In both studies, patients were screened for human immunodeficiency virus (HIV) infection and a positive HIV diagnosis was an exclusion criterion. In the PKDL study, patients were given paromomycin 20 mg/kg/day intramuscularly for 14 days and oral miltefosine (allometric dosing) for 42 days (PM/MF arm) or liposomal amphotericin B (AmBisome) 5 mg/kg/day intravenously at days 1, 3, 5, and 7 and oral miltefosine (allometric dosing) for 28 days (LAmB/MF arm). In the VL study, patients were given paromomycin 20 mg/kg/day intramuscularly for 14 days combined with oral miltefosine (allometric dosing) for 14 days or 28 days (PM/MF 14D and PM/MF 28D arms). Miltefosine allometric dosing was calculated according to patient's weight, height, and sex, and was applied to patients weighing <30 kg [15]. For patients weighing ≥30 kg, the allometric dose corresponded to the conventional dose of 2.5 mg/kg/day, with a maximum 150 mg/day. Therefore, patients weighing ≥30 to 44 kg received 100 mg/day and patients ≥45 kg received 150 mg/day. Details regarding the sampling schedule and bioanalysis assays are documented in the Supplementary Methods (S1).

Paromomycin Population PK Analysis

Previous population PK studies in VL patients in India and Eastern Africa demonstrated a decrease in paromomycin clearance over time [16, 17]. This time-varying kinetics observed in paromomycin is most likely associated with megalin-mediated tubular reabsorption. Aminoglycosides are primarily eliminated through glomerular filtration, with a fraction of the drug accumulating in the proximal tubule via the megalin receptors on the apical surface [18, 19]. Following binding and uptake by megalin, aminoglycosides undergo endocytosis and transfer into lysosomes, where they accumulate and may reach a critical threshold for release [20]. This reabsorption process is saturable and is associated with the development of kidney injury [20].

The renal clearance of paromomycin comprises glomerular filtration of the free drug fraction, tubular reabsorption, and tubular secretion processes, given by equation 1.

$$C{L}_R=fu*GFR-C{L}_{reabsorption}+C{L}_{secretion}$$ (1)

where CL_R is the total renal drug clearance, fu is the fraction of the drug unbound to plasma proteins, GFR is the glomerular filtration rate, CL_reabsorption is drug reabsorption from the renal tubule, and CL_secretion is the drug clearance by tubular secretion. Assuming free paromomycin is solely cleared through glomerular filtration, CL_secretion could be omitted. Additionally, CL_reabsorption was assumed to return to the central compartment without being excreted from the renal tubule.

Since the true GFR is unknown, renal drug clearance was informed by an estimate of GFR based on serum creatinine (eGFR_SCr) measured at baseline and on day 14. For individuals aged ≥18 years, the race-neutral Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) formula (2021) was utilized [21], and for those <18 years, the bedside Schwartz formula was used [22]. However, the existing formulas for calculating eGFR_SCr do not consider variability in creatinine production rates [23], which might play an important role in malnourished/severely ill populations like PKDL and VL patients. To evaluate this, eGFR_SCr was further adjusted by factors specific to VL and PKDL and was calculated to an absolute individual value by adapting the outcome of these formulas to the individual body surface area (BSA), as denoted in equation 2.

$$eGF{R}_{adjusted}=eGF{R}_{SCr}*{\displaystyle \frac{BSA}{1.73}*{\theta }^{PKDL}*{\theta }^{VL}}$$ (2)

in which adjustment factors θ^PKDL and θ^VL were estimated to describe the fractional difference between eGFR_SCr and eGFR_adjusted in PKDL and VL patients, respectively. This eGFR_adjusted was further related to CL_R using equation 3, which is a simplification of equation 1.

$$C{L}_R=(fu*eGF{R}_{adjusted})*(1-FR)$$ (3)

Here, fu was fixed to 0.67 based on protein binding of paromomycin in literature [11], and FR represents the fraction of drug reabsorbed from the renal tubules, accounting for the negative contribution to clearance. CL_reabsorption is derived from equation 4, in which the saturable reabsorption process, mediated by megalin receptor binding, was characterized using a maximal binding capacity (B_max) model represented by equation 5.

$$C{L}_{reabsorption}=(fu*eGF{R}_{adjusted})*FR*\left(1-{\displaystyle \frac{A_{tubular}}{B_{max}}}\right)$$ (4)

$${\displaystyle \frac{d{A}_{tubular}}{dt}=\left({\displaystyle \frac{C{L}_{reabsorption}}{V_C}}\right)*A_{Central}*-k_{out}*A_{tubular}}$$ (5)

In equation 4, B_max represents the maximal capacity for drug accumulation in the tubular compartment and A_tubular is the drug amount in this compartment at given time point. In equation 5, the first-order rate constant of megalin-mediated reabsorption was obtained by dividing the CL_reabsorption by the volume of distribution of the central compartment (V_C) and k_out describes the first-order rate constant for drug release from the tubular compartment to the central compartment. Factors that might affect the amount of megalin receptors and the drug-megalin interaction, including severity of the disease (VL relative to PKDL), body weight, age, sex, z-score, and albumin, were evaluated as covariates on B_max.

Miltefosine Population Pharmacokinetic Analysis

The population PK of miltefosine in VL patients from Eastern Africa has been relatively well studied, and we used previously developed models as a starting point [13–15]. Two covariates on relative oral bioavailability (F) found in previous models were evaluated sequentially. First, it was tested whether disease-specific differences were present in the previously observed reduced F during the first week of treatment. Second, the potential effect of cumulative miltefosine dose on F was examined [13–15]. Additionally, since miltefosine is prone to interact with lipids, the potential interaction of LAmB with miltefosine was evaluated in PKDL patients. Details regarding the miltefosine structural model and covariate analysis are documented in the Supplementary Methods (S2).

Exposure and Target Attainment

Secondary PK parameters for paromomycin and miltefosine were calculated using the individual estimates of the final PK models. Paromomycin exposure was determined by the area under the plasma concentration-time curve (AUC) from treatment start to day 1 (AUC_0-D1), day 14 (AUC_0-D14), and infinity (AUC_0-inf). Miltefosine exposure was expressed by the AUC from start until day 7 (AUC_0-D7), last dosing day (AUC_0-Dlast), and infinity (AUC_0-inf). The time point at which the miltefosine concentration reached the in vitro susceptibility value EC₉₀ (TEC₉₀) and the total time period that miltefosine concentration was above the EC₉₀ (Time >EC₉₀) were calculated. The EC₉₀ value of 10.6 mg/L was selected based on intracellular amastigote in vitro susceptibility testing of Leishmania donovani [14].

Analysis Software and Model Evaluation

Population PK analysis was performed using the nonlinear mixed-effects modeling program NONMEM (version 7.5; ICON Development Solutions, Ellicott City, Maryland), aided by Perl-speaks-NONMEM (version 5.0) and Pirana (version 2.9.9) for run deployment [24, 25]. Model parameters were estimated using the first-order conditional estimation with interaction method. Interindividual variability in PK parameters was estimated with an exponential variance model and residual unexplained variability was estimated with a proportional error model. Individual PK parameters were obtained by maximum a posteriori Bayesian estimation using POSTHOC option of NONMEM.

Model adequacy was guided by physiological plausibility, statistical significance, and graphical evaluation. The change in objective function value, which equals minus 2 times the log-likelihood, was used to define statistical significance between hierarchical models following a χ² distribution with 1 degree of freedom. A decrease in objective function value of ≥3.84, representing a P value of <.05, was considered statistically significant. Goodness-of-fit plots were used to assist graphical evaluation using R (version 4.0) and Xpose (version 4). A visual predictive check and sampling importance resampling were performed to assess the predictive performance and the parameter precision with 95% confidence intervals (CIs) for the final model [26, 27].

RESULTS

Patients and Data

Treatment details, baseline characteristics, and laboratory examination results of the included patients are presented in Tables 1, 2, and 3. Plasma concentration-time profiles of paromomycin and miltefosine are depicted in Supplementary Figures 1 and 2, respectively. A total of 328 paromomycin observations and 1516 miltefosine observations were collected in 109 PKDL and 264 VL patients (Table 1). The majority of the patients were children and adolescents aged ≤18 years, accounting for 96% and 72% of included PKDL and VL patients, respectively (Table 2). Severely malnourished patients were excluded from both clinical trials. For both VL and PKDL, the pediatric populations were mildly malnourished, as reflected by z-scores with interquartile range between −2.0 and −0.5 (Table 2). As both populations were from extremely poor areas and were to a large extent malnourished, body mass index (BMI) was not considered as an indicator for disease status. The distribution of body weight and BMI was similar between PKDL and VL patients for the different age groups, indicating comparable body size (Table 2). The median albumin level and neutrophil count at baseline were 0.71- and 0.34-fold lower, respectively, in VL versus PKDL patients, while the creatinine level was 2-fold higher (Table 3), confirming expected disease-specific differences in these hematological and biochemical variables.

Table 1.

Patient Treatment Details

Population	PKDL	VL
Total No. of patients	109	264
Included patients in the PK analysis, No.
PM model	14	26
MF model	109	264
PKDL treatment regimen, No. (%)
LAmB + MF 28 d	54 (49.5)	…
PM + MF 42 d	55 (50.5)	…
VL treatment regimen, No. (%)
PM + MF 28 d	…	167 (63.3)
PM + MF 14 d	…	97 (36.7)

Abbreviations: LAmB, lipsosomal amphotericin B; MF, miltefosine; PK, pharmacokinetic; PKDL, post-kala-azar dermal leishmaniasis; PM, paromomycin; VL, visceral leishmaniasis.

Table 2.

Patient Baseline Characteristics

	PKDL			VL
	Age ≤10 (n = 83)	10 < Age ≤18 (n = 22)	Age >18 (n = 4)	Age ≤10 (n = 130)	10 < Age ≤18 (n = 61)	Age >18 (n = 73)
Age, y	8.00 (6.00–9.00)	12.0 (11.0–13.8)	25.5 (22.8–27.8)	8.00 (6.00–9.75)	13.0 (11.0–15.0)	23.0 (21.0–28.0)
Body weight, kg	21.0 (18.0–24.8)	35.0 (31.0–40.8)	63.5 (55.5–71.0)	20.5 (17.1–25.0)	35.0 (32.0–41.0)	51.0 (47.0–54.0)
Height, cm	121 (113–129)	148 (142–157)	178 (174–183)	123 (113–132)	149 (140–158)	170 (165–176)
BMI, kg/m2	14.3 (13.6–15.0)	15.8 (15.2–16.8)	20.0 (18.4–21.2)	13.5 (13.1–14.4)	15.6 (14.9–17.0)	17.8 (16.5–18.7)
Body surface area, m2	0.87 (0.74–0.95)	1.22 (1.12–1.37)	1.79 (1.67–1.92)	0.86 (0.74–0.98)	1.21 (1.10–1.35)	1.58 (1.51–1.66)
Fat-free mass, kg	17.4 (14.7–20.1)	28.3 (24.7–35.2)	53.4 (48.3–58.5)	17.1 (14.3–20.7)	29.0 (25.9–35.5)	44.9 (42.5–48.0)
z-scorea	−1.19b (−1.84, −0.58)	−1.18c (−1.55, −0.56)	…	−1.04b (−1.78, −0.52)	−1.63c (−2.04, −1.02)	…
Male sex, No. (%)	47 (56.6)	18 (81.8)	4 (100)	89 (68.5)	52 (85.2)	72 (98.6)

Data are presented as median (interquartile range) unless otherwise indicated.

Abbreviations: BMI, body mass index; PKDL, post-kala-azar dermal leishmaniasis; VL, visceral leishmaniasis.

^aThe z-scores were calculated using the World Health Organization (WHO) Child Growth Standards for children aged between zero and 60 months or the WHO Growth References for school-aged children and adolescents using R package “zscorer.”

^bFor children aged below and equal to 10 years, z-scores were calculated based on weight-for-age.

^cFor children aged between 10 and 18 years, z-scores were calculated based on BMI-for-age.

Table 3.

Laboratory Examinations

Population	PKDL (n = 109)	VL (n = 264)	P Valuea
Albumin, g/L	38.8 (36.5–41.4)	27.4 (23.3–32.4)	<.001
Creatinine, mg/dL
Baseline	0.4 (0.2–0.5)	0.8 (0.6–1.0)	<.001
Day 14	0.6 (0.5–0.8)	0.8 (0.6–1.1)b	<.001
Absolute eGFR, mL/minc
Baseline	78.5 (55.2–108)	56.6 (33.1–98.3)	<.001
Day 14	50.3 (34.2–65.9)	50.2 (30.8–90.2)b	.597
Absolute neutrophil, ×103/µL
Baseline	2.91 (2.45–3.56)	1.00 (0.700–1.40)	<.001
Day 14	3.42 (2.70–4.32)	2.20 (1.70–2.80)	<.001

Data are presented as median (interquartile range) unless otherwise indicated.

Abbreviations: eGFR, estimated glomerular filtration rate; PKDL, post-kala-azar dermal leishmaniasis; VL, visceral leishmaniasis.

^aThe Mann-Whitney U test was used to evaluate if there is a statistically significant difference between the medians of 2 groups. A P value ≤.05 indicates a significant difference in the median between the groups.

^bData were available in 262 patients.

^cThe eGFR is calculated from the race-neutral Chronic Kidney Disease Epidemiology Collaboration formula (2021) for age ≥18 years and bedside Schwartz formula for age <18 years.

Pharmacokinetics of Paromomycin in PKDL and VL

Plasma concentration-time profiles of paromomycin were best described by a 3-compartment model with a saturable compartment, characterizing renal clearance processes, as depicted in Figure 1. The overall clearance from the central compartment was subsequently divided into 2 fractions: CL_R representing total renal clearance (equation 3) and CL_reabsorption representing saturable tubular reabsorption (equation 4). The parameter estimates of the final model are summarized in Table 4.

Figure 1.

Structure of the final paromomycin model. The overall clearance from the central compartment was divided into 2 components: total renal clearance and tubular reabsorption, where the fraction of reabsorption is denoted as FR. eGFR_adjusted represents creatinine-based estimates of glomerular filtration rate adjusted for disease-specific creatine production. Abbreviations: B_max, maximal drug accumulation in the saturable tubular compartment; CL_R, total renal clearance; CL_reabsorption, tubular reabsorption; fu, fraction unbound to plasma protein; k_a, rate of absorption; k_out, first-order rate constant for drug release from tubule to central compartment; Q/F, intercompartmental clearance; Vc/F, volume of the central compartment; Vp/F, volume of the peripheral compartment.

Table 4.

Paromomycin Final Model Parameter Estimates

Paromomycin Fraction Unbound	Unit	Estimate	(95% CI)
Rate of absorption (ka)	1/h	2.5	(2.05–3.33)
Fraction tubular reabsorption (FR)	…	0.3	(.2–.39)
Total renal clearance (CLR)a	L/h	2.3	…
Reabsorption from renal tubules (CLreabsorption)a	L/h	0.98	…
Volume of central compartmentb (Vc/F)	L	6.92	(6.55–7.3)
Intercompartmental clearanceb (Q/F)	L/h	0.12	(.10–.15)
Volume of peripheral compartmentb (Vp/F)	L	20.2	(7.6–38.7)
Rate constant for drug release from renal tubules (kout)	L/h	0.009	(.005–.015)
Maximal drug accumulation in renal tubules (Bmax)	mg	85	(50–134)
Fold difference in Bmax (VL relative to PKDL)	…	0.55	(.41–.74)
Factor adjusting creatinine-based eGFR to renal clearancec
In PKDL	…	1 (fixed)	…
In VL	…	1.44	(1.23–1.71)
Between-subject variability in Vc/F	CV%	11	(6–15)
Between occasion variability in adjusted eGFR	CV%	40	(35–49)
Residual proportional error	CV%	17	(15–19)

Abbreviations: CI, confidence interval; eGFR, estimated glomerular filtration rate; PKDL, post-kala-azar dermal leishmaniasis; VL, visceral leishmaniasis.

^aCL_R and CL_reabsorption were calculated by fu*eGFR_adjusted*(1-FR) and fu* eGFR_adjusted *FR, respectively, using individual Empirical Bayes estimates of the final model. The values are presented as median.

^bAllometric scaled based weight with power exponent of 0.75 for clearance and 1 for volume of distribution. Estimate is given for a standardized body weight of 25 kg.

^cCreatinine-based eGFR is calculated from the race-neutral Chronic Kidney Disease Epidemiology Collaboration formula (2021) for age ≥18 years [21], and the bedside Schwartz formula for age <18 years [22].

Variabilities in CL_R between individuals and throughout the 14-day treatment period were adequately described by eGFR_adjusted, where the disease-specific adjustment θ^PKDL was estimated and subsequently fixed to 1, while θ^VL was estimated at 1.44 (95% CI, 1.23–1.71), accounting for differences in creatinine production. This result suggested that approximating CL_R of paromomycin using solely serum creatinine-based eGFR is feasible in PKDL patients, while this may lead to underestimation of CL_R in VL patients by 44%.

Incorporating a saturable compartment to reflect paromomycin reabsorption via megalin significantly improved the model fit and explained the time-dependent trends (details in Supplementary Methods [S3]). In the final model, FR was estimated at 0.3 (95% CI, .2–.39), the maximal first-order rate constant for megalin-mediated drug reabsorption was 0.13 L/hour, and k_out for drug release into central compartment was estimated at 0.009 L/hour (95% CI, .005–.015). Severity of the disease was a significant covariate on B_max, indicating that the maximum capacity for paromomycin accumulation in the tubular compartment was 0.55-fold (95% CI, .41–.74) lower in VL patients compared to PKDL patients. No additional significant covariates were found to explain variability in B_max. The final model adequately described paromomycin observations in both PKDL and VL patients. Goodness-of-fit plots and the VPC based on this model showed no major deviations (Supplementary Figures 3 and 4).

Pharmacokinetics of Miltefosine in PKDL and VL

In line with previous studies, miltefosine concentration-time profiles were best described by a 2-compartment model with first-order absorption and first-order elimination. The parameter estimates of the final model are summarized in Table 5. Nonlinearities in miltefosine PK were identified and were best described by 3 effects on F: disease-related reduction in the first week (COV_{F, W1}), cumulative dose (COV_{F, CD}), and a drug–drug interaction with LAmB (COV_{F, LAmB}). Details are further elaborated in the Supplementary Methods (S4).

Table 5.

Miltefosine Final Model Parameter Estimates

Parameter	Unit	Estimate	(95% CI)
Absorption rate in PKDL (ka PKDL)	L/day	5.43	(4.54–6.26)
Absorption rate in VL (ka VL)	L/day	0.99	(.84–1.17)
Clearancea (CL/F)	L/day	1.53	(1.4–1.66)
Volume of central compartmenta (V2/F)	L	13.8	(12.66–14.89)
Intercompartmental clearance (Q/F)	L/day	0.039	(.032–.049)
Volume of peripheral compartmenta (V3/F)	L	2.01	(1.69–2.54)
F after a week of treatment	%	100 (fixed)	…
Relative decrease in F during the first week of treatment (COVF,W1)
PKDL patients	%	0 (fixed)	…
VL patients	%	69	(62–76)
Coadministration of LAmB (COVF,LAmB)	%	8.6	(2–15)
Exponent of power relationship between cumulative MF dose and F (COVF,CD)	…	−0.13	−.16 to −.11
Between-subject variability (IIV)
CL/F	CV%	20	(17–22)
V2/F	CV%	13b	…
COVF,W1 in VL patients	CV%	71	(61–83)
Residual proportional error	CV%	31	(30–33)

Abbreviations: CI, confidence interval; F, relative bioavailability; IIV, interindividual variability; LAmB, liposomal amphotericin B; MF, miltefosine; PKDL, post-kala-azar dermal leishmaniasis; VL, visceral leishmaniasis.

^aAllometric scaled based on fat-free mass with power exponent of 0.75 for clearance and 1 for volume of distribution. Estimate is given for a standardized fat-free mass of 25 kg.

^bIIV in V2/F was not unidentifiable, but high correlation between CL/F and V2/F was suggested by sampling importance resampling result. Therefore, the same variance was assumed for IIV in CL/F nd V2/F, and a scaling factor of 0.45 was estimated.

In the concentration-time profiles, a large difference in drug accumulation and exposure between VL patients and PKDL patients was observed in the first week of treatment (Supplementary Figure 2). In VL patients, F in the first week of treatment decreased by 69% (95% CI, 62%–76%). However, in PKDL patients, a decrease in F was not identified in the first week of treatment; therefore, COV_{F, W1} was fixed to 0 (Supplementary Figure 5). In addition to the difference in F, the rate of absorption (k_a) in PKDL and VL was 5.43 day⁻¹ (95% CI, 4.54–6.26) and 0.99 day⁻¹ (95% CI, .84–1.17), respectively, indicating a much longer absorption half-life in VL patients. Simulated concentration-time profiles for a typical PKDL patient and a VL patient receiving a 28-day miltefosine regimen are presented in Figure 2. Since PKDL patients are not affected by a systemic parasitic infection, the relatively low k_a and F at treatment start found only in VL patients are consistent with previous hypotheses about systemic VL infection–related malabsorption [28, 29]. Despite a similar daily dosing regimen, in the first week median miltefosine concentration was 12% lower in PKDL patients who received LAmB-MF than in patients receiving PM-MF (Supplementary Figure 2). LAmB was found to reduce miltefosine's F by 8.6% (95% CI, 2%–15%) in the first week. The final model adequately described miltefosine observations in both PKDL and VL patients. Goodness-of-fit plots and the VPC based on this model showed no major deviations (Supplementary Figures 6 and 7).

Figure 2.

Simulated concentration-time profiles for a typical post-kala-azar dermal leishmaniasis (PKDL) patient (red line) and a visceral leishmaniasis (VL) patient (blue line) with a fat-free mass of 25 kg receiving a 28-day miltefosine regimen. The dashed line represents end of miltefosine treatment (day 28). Miltefosine concentrations during the first week of treatment were lower in VL patients but approached levels similar to those in PKDL patients after the third week of treatment.

Exposure and Target Attainment in PKDL and VL

Plasma exposures and target attainment of paromomycin and miltefosine are summarized in Tables 6 and 7, respectively. In the populations studied, median paromomycin AUC_0-D1 and AUC_0-D14 in PKDL patients was 0.74- and 0.86-fold lower than in VL patients, respectively. Miltefosine AUC_0-D7 irrespective to treatment arms was 3.7-fold higher in PKDL patients compared to VL patients (median, 98.6 mg*day/L in PKDL and 26.1 mg*day/L in VL). In the 28-day miltefosine treatment arms, AUC_0-D28 was 1.4-fold higher in PKDL patients compared to VL patients. Due to differences in initial drug accumulation, TEC₉₀ was 2 and 9 days after the treatment start in PKDL and VL patients, respectively.

Table 6.

Paromomycin Plasma Exposure and Target Attainment

Parameter	PKDL (n = 14)	VL (n = 26)
AUC0-D1 (mg*h/L)	127 (117–148)	171 (124–207)
AUC0-D14 (mg*h/L)	2370 (2130–2420)	2770 (2380–3440)
AUC0-inf (mg*h/L)	2460 (2220–2500)	2840 (2430–3520)

Data are presented as median (interquartile range). Both PKDL and VL patients received 14 days of 20 mg/kg paromomycin.

Abbreviations: AUC_0-D1, area under the plasma concentration-time curve from start until day 1; AUC_0-D14, area under the plasma concentration-time curve from start until day 14; AUC_0-inf, area under the plasma concentration-time curve from start until time infinity; PKDL, post-kala-azar dermal leishmaniasis; VL, visceral leishmaniasis.

Table 7.

Miltefosine Exposure and Target Attainment

Treatment Regimen	PKDL		VL
	LAmB + 28-Day MF (n = 54)	PM + 42-Day MF (n = 55)	PM + 14-Day MF (n = 167)	PM + 28-Day MF (n = 97)
AUC0-D7 (mg*day/L)	95.6 (89.9–103)	107 (74.4–159)	28.2 (21.3–34.9)	24.9 (22.1–27.6)
AUC0-Dlast (mg*day/L)	720 (658–762)	1130 (1010–1260)	129 (113–154)	514 (452–548)
AUC0-inf (mg*day/L)	991 (885–1090)	1410 (1260–1570)	352 (301–415)	808 (693–906)
TEC90 (days)	2.10 (2.01–2.30)	2.06 (1.95–2.17)	8.69 (7.82–9.33)	9.11 (8.78–9.59)
Time >EC90 (days)	34.4 (33.2–35.7)	48.1 (46.5–50.1)	12.7 (10.9–14.9)	28.7 (26.6–31.2)

Data are presented as median (interquartile range).

Abbreviations: AUC_0-D7, area under the plasma concentration-time curve from start until day 7; AUC_0-Dlast, area under the plasma concentration-time curve from start until the end of last dosing day; AUC_0-inf, area under the plasma concentration-time curve from start until time infinity; EC₉₀, 90% effective miltefosine concentration, equivalent to 10.6 µg/mL; LAmB, liposomal amphotericin B; MF, miltefosine; PKDL, post-kala-azar dermal leishmaniasis; PM, paromomycin; TEC₉₀, time point that the miltefosine plasma concentration reached the in vitro susceptibility value EC₉₀; Time >EC₉₀, total time period that the miltefosine concentration was above the in vitro susceptibility value EC₉₀ from the treatment start until time infinity; VL, visceral leishmaniasis.

DISCUSSION

Large and clinically relevant differences in drug exposure and PK parameters for paromomycin and miltefosine between PKDL and VL patients in Eastern Africa were identified and quantified for the first time. In PKDL patients, the AUC_0-D1 of paromomycin was 0.74-fold lower, while AUC_0-D28 of miltefosine was 1.4-fold higher compared to VL patients. These considerable differences in exposure indicate that dosing regimens cannot be directly extrapolated from VL to PKDL, especially in Eastern Africa.

The reduced plasma exposure to paromomycin in PKDL patients is associated with greater drug accumulation in the renal tubule and higher renal clearance, attributed to better renal function at baseline. This lower systemic exposure could potentially account for the lower occurrence of ototoxicity in PKDL patients. For instance, ear and labyrinth disorders were observed in 5.3% of VL patients but were absent in PKDL patients in the clinical trials [7, 9]. While the current paromomycin-based regimen was shown to be effective in managing PKDL patients from eastern Africa, this suggests that higher paromomycin dosages might be tolerable in PKDL patients compared to VL patients [9]. On the other hand, the higher F for miltefosine found in PKDL patients is likely due to better gastrointestinal absorption, which resulted in increased miltefosine exposure and faster achievement of EC₉₀. This, in turns, implies the possibility of a shorter miltefosine treatment duration in PKDL patients from Eastern Africa.

In the paromomycin model, the time-dependent exposure observed was explained by alterations in renal function, partly influenced by saturable tubular reabsorption. After glomerular filtration, aminoglycosides are reabsorbed through megalin receptors on the apical tubular surface and accumulate in cell lysosomes, which can result in further nephrotoxicity [18–20]. In VL patients, B_max was 0.55-fold (95% CI, .41–.73) lower than in PKDL patients, explaining the slower decrease in paromomycin concentration observed within 4 hours postdose. Renal involvement, commonly observed in VL [12], could contribute to this lower reabsorption capacity found in VL patients, as symptoms like albuminuria can signify tubular stress and decreased activity of megalin, leading to decreased tubular reabsorption [30]. Although it remains unclear whether aminoglycosides’ renal toxicity is related to local drug concentration or drug release from lysosomes [18–20], the lower reabsorption capacity in VL patients might suggest a reduced tolerance for aminoglycoside-induced renal toxicity. The framework of the current model can be extended to other aminoglycosides to further explore the relationship between drug accumulation in renal tubules and drug-induced toxicity.

A decrease in the renal clearance of paromomycin was observed during the 14-day treatment period in both PKDL and VL patients. In PKDL patients, the renal clearance aligned with creatinine-based eGFR, whereas in VL patients, correlating renal clearance with creatinine-based eGFR required a disease-specific correction of 1.44-fold. This discrepancy suggests that the measured serum creatinine level did not accurately reflect the renal function, particularly in this severely ill and malnourished population. This could be attributed to a lack of validated equations for estimating GFR based on serum creatinine in this particular population, wherein the deviating production of creatinine is not accurately accounted for [23]. The validity of current creatinine-based eGFR estimation methods in similar populations has been questioned previously [31].

For miltefosine, F was decreased by 69% (95% CI, 62%–76%) in VL patients in the first week of treatment, while in PKDL patients no initial decrease in F was noticeable. The k_a in VL patients was 0.99 day⁻¹ (95% CI, .84–1.17), comparable to previously reported values of 0.4–1.6 day⁻¹ [13–15]. The k_a in PKDL patients was described for the first time, equal to 5.43 day⁻¹ (95% CI, 4.54–6.26), which was much faster than the VL patients, but comparable to the k_a of 8.64 day⁻¹ reported for patients with cutaneous leishmaniasis [32]. The reduced F and slower k_a imply poor gastrointestinal absorption in VL patients, probably owing to a more severe disease status or, potentially, a local gastrointestinal effect of the systemic Leishmania infection [33].

Coadministration of LAmB was found to lower miltefosine exposure by around 10%, suggesting a potential drug–drug interaction between miltefosine and LAmB. The decreased miltefosine exposure could potentially be associated with the higher frequency of vomiting observed in the LAmB/MF arm of the PKDL trial, which mainly occurred during the first 2 weeks of treatment, even though the drug was readministered in case of vomiting [9]. Beside this, free or micellular miltefosine can theoretically be incorporated into the liposomes of LAmB, potentially reducing systemic exposure [34, 35]. However, this additional miltefosine clearance route might actually be beneficial for increased miltefosine target-site delivery and activity, as liposomes are preferentially phagocytized by parasite-infected macrophages [36].

This study illustrates and highlights the difficulty of directly extrapolating dosing regimens between VL and the various other dermal clinical presentations of leishmaniasis, particularly for orally administered and renally cleared drugs. Moreover, various disease-specific mechanisms affecting the absorption, distribution, and elimination of paromomycin and miltefosine were identified and characterized. These findings should be taken into consideration when developing new treatments for leishmaniasis, emphasizing the importance of characterizing PK and PK/PD relationships across different clinical manifestations of the disease. To optimize dosing of paromomycin and miltefosine for PKDL patients, it is essential to further establish exposure-response relationships, with a focus on the relevance of drug exposure within skin target site.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online (http://jid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.

Notes

Acknowledgments. The authors thank the patients involved in this study and their families and communities for their willingness to participate in the trial; all co-investigators, nurses, laboratory personnel, and hospital administrators who allowed the authors to conduct the study in their respective study sites; and staff at the 5 Leishmaniasis East Africa Platform (LEAP) sites and 2 Doctors Without Borders (Médecins Sans Frontières [MSF]) sites: Kacheliba in Kenya; Amudat in Uganda; Doka, Umelkher, and Tabarakallah (MSF) in Sudan; and Gondar and Abdurafi (MSF) in Ethiopia. The Drugs for Neglected Diseases initiative also thanks UK aid, MSF International, and the Swiss Agency for Development and Cooperation for supporting its overall mission.

Financial support. This work was supported by the European and Developing Countries Clinical Trials Partnership (EDCTP2) program supported by the European Union (grant number RIA2016S-1635 AfriKADIA); the Dutch Ministry of Foreign Affairs; the Federal Ministry of Education and Research through KfW, Germany; Medicor Foundation; Médecins sans Frontières International; Swiss Agency for Development and Cooperation, Switzerland; UK aid, United Kingdom; and other private individuals and foundations. T. P. C. D. was supported by the Dutch Research Council (NWO/ZonMw; project 91617140) and the Swedish Research Council (VR 2022–01251).